Comparative persistence and adherence to overactive bladder medications in patients with and without diabetes

Authors


  • Disclosures Stephen Johnston, David Smith and Kathleen Wilson are employees of Thomson Reuters, which was paid by Astellas Pharma Global Development, Inc., to conduct this study. Stephen Janning is an employee of GlaxoSmithKline. Gabriel Haas, Stan Bukofzer, GinaMarie Reckard and Shun-Ping Quan are employees of Astellas Pharma Global Development, Inc.

Gabriel P. Haas
Astellas Pharma Medical Affairs, Three Parkway North, Deerfield, IL, USA
Tel.: +1 8473178994
Fax: +1 8473171275
Email: gabriel.haas@us.astellas.com

Summary

Aims:  This retrospective administrative claims-based study evaluated comparative persistence and adherence to overactive bladder (OAB) medications in US patients with and without diabetes.

Methods:  Patients ≥ 18 years who initiated OAB medications between 1 January 2005 and 30 June 2008 were analysed from the Truven Health MarketScan® Commercial and Medicare Supplemental databases. A 12-month baseline period prior to OAB medication initiation was used to classify patients into diabetes and non-diabetes cohorts, and measure demographic and clinical characteristics. Patients in each cohort were directly matched 1 : 1 based on index year, age, gender and geographic region. Multiple logistic regression was used to compare cohorts on outcomes of ≥ 80% adherence to OAB medications and refilling a second OAB medication prescription. Cox’s proportional hazards model compared time to non-persistence with OAB medications between both cohorts.

Results:  In total, 36,560 patients were included in each cohort. Compared with the non-diabetes cohort, the diabetes cohort had 21.5% higher odds of ≥ 80% adherence to OAB medications, 16.6% higher odds of filling a second OAB medication prescription and 10.3% lower hazard of non-persistence with OAB medications during a 12-month evaluation period.

Conclusions:  Patients with diabetes were more persistent and adherent to OAB medications and had higher odds of filling a second medication prescription than patients without diabetes. Further research is needed to identify factors responsible for these findings.

What’s known

Bladder-related complications are common in patients with diabetes mellitus. Poor adherence and persistence to therapy in such patients can result in recrudescence of symptoms and high retreatment rates.

What’s new

We performed a retrospective administrative claims-based analysis to evaluate comparative adherence and persistence to overactive bladder (OAB) medications in patients with and without diabetes. We found that patients with diabetes display significantly better adherence and persistence to OAB medication. Physicians managing these patients should be aware of the importance of this association for the purposes of diagnosis and management.

Introduction

Overactive bladder (OAB) is a ‘syndrome of symptoms’, defined by the International Continence Society as urinary urgency, with or without urgency incontinence, usually accompanied by increased daytime urinary frequency and nocturia (1). It is a highly prevalent disorder, with considerable impact on health-related quality of life. Estimates of the prevalence of OAB from population-based studies range from 7% to 27% in men and 9% to 43% in women (2–4). A landmark US epidemiology study reported the incidence of OAB to be ∼16–17% for both genders (5), which corresponds to around 33 million adults in the USA. Lower urinary tract symptoms that include OAB are common in older individuals with diabetes mellitus and can affect work productivity, sexual satisfaction and overall health (6). Diabetes-associated bladder complications can be associated with peripheral nerve irritation, detrusor overactivity and an increase in bladder sensation (7). A recent study found that ∼23% of patients with diabetes also had OAB; common OAB symptoms included nocturia (48%), frequency (46%), urgency (23%) and urge incontinence (13%) (8).

Adherence is defined as the extent to which a patient takes his or her medication at the correct dosage and interval freely and purposefully (9–11), whereas persistence is defined as ‘the duration of time from initiation to discontinuation of therapy’ or the extent to which the patient’s behaviour matches the prescriber’s recommendations with respect to timing, dosing and frequency (10,12). In general, patients with poor medication adherence have increased morbidity (13,14), which can be associated with increased healthcare costs (11,15). Good persistence and adherence to antimuscarinic therapy is an important goal as discontinuation of, or incomplete adherence to, OAB medication could result in symptom relapse and an increase in retreatment rates (16). Despite the proven efficacy of OAB medications in reducing symptoms related to OAB, side effects associated with this class of drugs, patient beliefs about OAB and its treatment, and cost may negatively affect adherence and persistence (17–19). Prior studies have demonstrated that within the first year of treatment, 80–90% of patients discontinue their OAB medications and only 30% of patients have an adherence rate of > 80% (20–22). Medication adherence is a complex phenomenon that is affected by behavioural, social, economic, ethical and psychological factors (17,23). Certain chronic and debilitating diseases require strict medication regimens and frequent encounters with healthcare providers to foster adherence. Patients with diabetes, for example, may be accustomed to maintaining multiple medication regimens (24), have frequent contact with physicians and, via specialist interventions, may be more aware of the importance of non-persistence and non-adherence to their medications (25,26).

Given the issues identified with adherence to OAB medications, there is a gap regarding our understanding of adherence and persistence in diabetic patients compared with the general population of OAB patients. The objective of the current study was to evaluate and compare persistence and adherence to OAB treatments in patients with and without diabetes. This study therefore sought to test the following hypotheses: (i) patients with diabetes who are initiating OAB medications will be more adherent to their OAB medications over a 12-month evaluation period than demographically matched patients without diabetes, (ii) patients with diabetes who are initiating OAB medications will be more likely to refill a second OAB medication prescription over a 12-month evaluation period than demographically matched patients without diabetes and (iii) patients with diabetes who are initiating OAB medications will have a longer time to non-persistence with OAB medications over a 12-month evaluation period than demographically matched patients without diabetes.

Methods

Data source

This was a retrospective analysis of administrative claims data. The Truven Health MarketScan® Commercial Claims and Encounters (Commercial) database and the Medicare Supplemental and Coordination of Benefits (Medicare) database (Truven Health, Ann Arbor, MI, USA) were used in combination for this study. The Truven Health MarketScan® Databases are constructed from privately insured, paid medical and prescription drug claims. The Commercial database contains the inpatient medical, outpatient medical, and outpatient prescription drug claims and encounter records of several million employees and their dependants, whereas the Medicare database contains the inpatient medical, outpatient medical, and outpatient prescription drug claims and encounter records of individuals with Medicare supplemental insurance. The medical claims are linked to outpatient prescription drug claims and person-level enrolment data. The data contained in the Commercial and Medicare databases are statistically de-identified and fully compliant with the Health Insurance Portability and Accountability Act (HIPAA) Privacy Regulations.

Study period and sample selection criteria

The study employed data from 1 January 2004 through 30 June 2009 (66-month period) that represented the latest available data in the Truven Health MarketScan® Commercial and Medicare Supplemental databases (Figure 1). The study period was subdivided into the patient selection period, baseline period and evaluation period. The patient selection period ran from 1 January 2005 through 30 June 2008 (42-month period). During this time, patients who were candidates for inclusion in the study sample were identified in the databases by searching each patient’s claims history for the first record (chronologically) of an outpatient prescription claim for one of the following OAB medications: darifenacin hydrobromide, oxybutynin chloride, solifenacin succinate, tolterodine tartrate or trospium chloride. This was used as a surrogate for OAB diagnosis.

Figure 1.

 Study period

The date of the first claim for the initial OAB medication was defined as the index date and the baseline period was 12 months prior to a patient’s index date (not including the index date). The baseline period was used to establish a ‘clean period’ during which there were no prescriptions for OAB medications and was used to identify the patient’s diabetes status, baseline demographics and clinical characteristics (Figure 1). The evaluation period was specific to each individual and comprised 12 months after the index date (including the index date). The evaluation period was used to measure and evaluate various aspects of a patient’s persistence and adherence to OAB medications.

The observation period was specific to each individual. Patients aged ≥ 18 years were included in the study sample if they had ≥ 1 outpatient OAB medication claim filled during the period from 1 January 2005 to 30 June 2008, and had continuous enrolment and pharmacy benefits throughout the 24-month baseline and evaluation periods. Note that OAB may be under-coded on administrative claims data, which means that patients may have initiated antimuscarinic therapy but that their physician may not actually record an International Classification of Diseases-Ninth Revision – Clinical Modification (ICD-9-CM) diagnosis code for OAB on a medical claim. Patients were not required to have a diagnosis of OAB during the study period for inclusion. Patients were excluded if they had a prescription for an OAB medication during the baseline period or a primary or secondary diagnosis of gestational diabetes, pregnancy or stress incontinence anytime during the study period.

Cohort assignment

In order to determine the association between diabetes status and OAB medication persistence and adherence, patients were classified into the diabetes cohort or the non-diabetes cohort. Patients were assigned to the diabetes cohort if they had ≥ 2 prescription claims for an antidiabetic medication (oral and/or injectable) during the baseline period, or had a primary or secondary diagnosis of diabetes mellitus (ICD-9-CM code 250.xx) during the baseline period; patients with insulin-dependent diabetes mellitus and non-insulin-dependent diabetes mellitus were included. Patients who met all sample inclusion and exclusion criteria except for criteria for the diabetes cohort were assigned to the pool of potential patients considered for the non-diabetes cohort.

In order to compare demographically similar groups of patients, non-diabetes cohort patients were matched to diabetes cohort patients via direct matching at a 1 : 1 ratio based on index year (year of first OAB prescription), age group (18–24, 25–34, 35–44, 45–54, 55–64, 65–74 and 75+ years of age), gender and US geographic region of residence.

Outcomes

Outcomes were OAB medication adherence, time to OAB medication non-persistence and refill of a second OAB medication prescription. OAB medication adherence was assessed using the interval-based (fixed time period) medication possession ratio (MPR). This was measured as the proportion of days with any OAB medication on hand, over the length of the 12-month evaluation period. For example, if during the 12-month evaluation period a patient filled 10 prescriptions, each with 30 days supplied, they would have been designated as having 300 days with OAB medication on hand, resulting in an MPR value of 82% (300/365 days). For the statistical analyses, this variable was dichotomised as a binary outcome = 1 if adherence ≥ 80%, and = 0 otherwise. In most studies, patients are considered non-adherent to medication if they do not take ≥ 80% of their prescribed drugs (11). This cut-off has also been used in previous OAB-specific retrospective analyses (18,22).

Time to OAB medication non-persistence was measured as the number of days from the index date until a gap in OAB medication of ≥ 45 days. The gap of ≥ 45 days was used in a previous analysis of OAB medication adherence and persistence, and accounts for the widespread intermittent use of OAB medications that has been found in previous studies (22). Refill of a second OAB medication prescription was a flag that indicated whether the patient ever filled a second OAB medication prescription at any point during the 12-month evaluation period.

OAB persistence and adherence variables were calculated based on OAB medications as a group and allowed for switches between OAB medications. As such, as long as patients were on any of the specified OAB medications, they were considered persistent/adherent.

Statistical analyses

Descriptive statistics were used to summarise the characteristics of patients in the diabetes vs. the non-diabetes cohort. Chi-squared tests were used to test for differences in nominal/categorical variables; t-tests were used to test for differences in interval/continuous variables. Statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC, USA).

Multivariate analyses were carried out to test the three study hypotheses. Multiple logistic regression was used to separately analyse the outcome of ≥ 80% adherence to OAB medications and the outcome of refilling a second OAB medication prescription. Cox’s proportional hazards model was used to compare the time to non-persistence with OAB medications in the diabetes vs. non-diabetes cohorts. Explanatory variables for these analyses included binary indicator for membership in the diabetes cohort, age, gender, insurance plan type, US Census Bureau geographic region of residence, employee relationship, urban/rural residence, total expenditures in the baseline period, the Deyo–Charlson Comorbidity Index (DCI), number of ICD-9-CM diagnosis codes, number of unique National Drug Codes (NDCs), number of inpatient admissions in the baseline period, baseline specialty visits, baseline OAB-related diagnoses, baseline medications, and baseline micro- and macro-vascular comorbidities. The DCI, number of ICD-9-CM diagnosis codes at the three-digit level of specificity, and NDC numbers were used to represent the baseline health of patients in both cohorts. Models were fitted to the data using stepwise regression as well as forced inclusion of the primary explanatory variable (diabetes cohort vs. non-diabetes cohort status). Variables were retained in the models if they had a p-value that was below or equal to the maximum p-value selection criterion of 0.05.

As patients who obtain 90-day OAB medication prescriptions introduce the potential to erroneously classify days supplied as representing periods of adherence, sensitivity analyses were conducted in which these patients were excluded from the analysis.

Results

Baseline demographics and characteristics

The database contained 58,090,659 patients, of whom 595,381 were prescribed ≥ 1 OAB medication in the time period between 1 January 2005 and 30 June 2008. The final study group comprised 197,423 patients, all of whom met the study inclusion criteria. Directly matching the diabetes cohort to the non-diabetes comparison cohort on index year, age group, gender and geographic region eliminated only one patient from the diabetes cohort (a match success rate of 99.997%), and produced matched diabetes and non-diabetes cohorts that each comprised 36,560 patients for a total of 73,120 patients for final inclusion in the study.

The demographic characteristics of the study sample, stratified by diabetes vs. non-diabetes cohorts, are shown in Table 1. As a result of direct matching on demographic characteristics, diabetes and non-diabetes cohort patients were demographically similar. In both cohorts, the mean age was approximately 69 years, 59.8% were female, and 77.7% resided in the North Central or Southern geographic regions.

Table 1.   Patient demographics
  Diabetes cohort (= 36,560) Non-diabetes cohort (= 36,560) p-Value
  1. SD, standard deviation.

Age (years), mean ± SD68.9 ± 12.069.1 ± 12.70.029
Gender, n (%)
 Male14,703 (40.2)14,703 (40.2)> 0.999
 Female21,857 (59.8)21,857 (59.8)> 0.999
Payer, n (%)  0.056
 Commercial13,077 (35.8)13,325 (36.4)0.056
 Medicare23,483 (64.2)23,235 (63.6)0.056
US Census geographic region, n (%)  1
 Northeast2938 (8.0)2938 (8.0)> 0.999
 North Central14,826 (40.6)14,826 (40.6)> 0.999
 South13,565 (37.1)13,565 (37.1)> 0.999
 West5145 (14.1)5145 (14.1)> 0.999
 Unknown86 (0.2)86 (0.2)> 0.999
Urbanicity, n (%)  0.957
 Urban29,755 (81.4)29,735 (81.3)0.849
 Rural6731 (18.4)6748 (18.5)0.871
 Unknown74 (0.2)77 (0.2)0.807

The baseline clinical measures, stratified by diabetes vs. non-diabetes cohort, are shown in Tables 2 and 3. As expected, patients in the diabetes cohort had poorer baseline health than those in the non-diabetes cohort, as indicated by higher mean values for various indices of health. Patients in the diabetes cohort were also more likely than those in the non-diabetes cohort to have various comorbidities (Table 2).

Table 2.   Baseline comorbidities
  Diabetes cohort (= 36,560) Non-diabetes cohort (= 36,560) p-Value
  1. OAB, overactive bladder.

Baseline OAB variables, n (%)
 Benign prostatic hyperplasia 319 (2.2)334 (2.3)0.553
 Bladder cancer755 (2.1)761 (2.1)0.876
 Interstitial cystitis 203 (0.6)282 (0.8)< 0.001
 Narrow angle glaucoma95 (0.3)81 (0.2)0.291
 Neurogenic bladder629 (1.7)533 (1.5)0.005
 Urinary or gastric retention1827 (5.0)1717 (4.7)0.058
Baseline diabetes-related variables, n (%)
 Type 2 diabetes mellitus31,035 (84.9)  
 Hypertension 19,494 (53.3)15,981 (43.5)< 0.001
 Acute myocardial infarction634 (1.7)339 (0.9)< 0.001
 Other ischaemic heart disease8977 (24.6)5795 (15.9)< 0.001
 Congestive heart failure3445 (9.4)1640 (4.5)< 0.001
 Cerebrovascular accident4105 (11.2)2765 (7.6)< 0.001
 Peripheral vascular disease4464 (12.2)2142 (5.9)< 0.001
 Diabetic peripheral neuropathy3418 (9.3)  
 Diabetic retinopathy 2076 (5.7)  
 Leg and foot amputation128 (0.4)22 (0.1)< 0.001
 Nephropathy217 (0.6)60 (0.2)< 0.001
Table 3.   Baseline clinical characteristics and healthcare utilisation
  Diabetes cohort (= 36,560) Non-diabetes cohort (= 36,560) p-Value
  1. ICD-9-CM, International Classification of Diseases-Ninth Revision – Clinical Modification; NDC, National Drug Codes; SD, standard deviation.

Number of preperiod hospitalisations, mean (%) 0.43 (0.01)0.29 (0.01)< 0.001
Number of unique 3-digit ICD-9-CM diagnosis codes, mean ± SD12.95 ± 0.0810.45 ± 0.07< 0.001
Number of unique NDCs, mean ± SD18.15 ± 0.1011.95 ± 0.08< 0.001
Total baseline expenditures, mean ± SD$21,548 ± 376$14,574 ± 246< 0.001
Deyo–Charlson Comorbidity Index, Mean ± SD2.2 ± 0.020.8 ± 0.01< 0.001
Number of preperiod outpatient visits, mean (SD)
 Primary care provider visit 4.84 (0.05)3.56 (0.04)< 0.001
 Endocrinology specialist care visit0.16 (0.01)0.02 (0.003)< 0.001
 Gynaecology specialist care visit among females0.29 (0.01)0.37 (0.01)< 0.001
 Urology specialist care visit among females0.31 (0.01)0.30 (0.01)< 0.001
 Urology specialist care visit among males1.39 (0.02)1.51 (0.02)< 0.001
 Other specialist care visit 4.98 (0.06)4.06 (0.05)< 0.001
Other baseline medications, n (%)
 Diuretics19,230 (52.6)12,341 (33.8)< 0.001
 Antispasmodics5267 (14.4)4585 (12.5)< 0.001
 Antihypertensive agents3972 (10.9)2761 (7.6)< 0.001
 Tricyclic antidepressants2589 (7.1)1983 (5.4)< 0.001
 Anti-Parkinsonian agents884 (2.4)904 (2.5)0.632

Patients in the diabetes cohort were more likely to have established contact or have had more frequent encounters with the healthcare system compared with those in the non-diabetes cohort, as indicated by higher mean number of hospitalisations, visits to a primary care provider, visits to an endocrinologist and mean healthcare expenditures during the baseline period. Female patients in the diabetes cohort had a slightly higher mean number of visits to urologists than those in the non-diabetes cohort, whereas the reverse was true for male patients. Female patients in the diabetes cohort also had a lower mean number of visits to gynaecologists compared with their counterparts in the non-diabetes cohort (Table 3).

Outcomes

Overactive bladder medication adherence

Unadjusted adherence and persistence outcomes, stratified by diabetes vs. non-diabetes cohort, are shown in Table 4. Patients in the diabetes cohort were more adherent to OAB medications than patients in the non-diabetes cohort. On average, 47.3% of the 12-month evaluation period was covered by OAB medications among patients in the diabetes cohort [interval-based MPR = 0.47; standard deviation (SD) = 0.004] compared with 42.4% of the evaluation period among patients in the non-diabetes cohort (interval-based MPR = 0.42; SD = 0.003); this difference was statistically significant (p < 0.001; Table 4).

Table 4.   Unadjusted adherence and persistence outcomes
  Diabetes cohort (= 36,560) Non-diabetes cohort (= 36,560) p-Value
  1. MPR, medication possession ratio; OAB, overactive bladder; SD, standard deviation.

Adherence (MPR), 12-month interval-based, mean ± SD0.473 ± 0.0040.424 ± 0.003< 0.001
Persistence in days, mean ± SD164.0 ± 1.4146.9 ± 1.4< 0.001
Patients filling a second OAB prescription of any type, n (%) 25,490 (69.7)23,595 (64.5)< 0.001
Reason for end of persistence, n (%)
 Discontinuation (45-day gap in therapy)26,144 (71.5)27,846 (76.2)< 0.001
 Censor because of reaching end of 12-month period10,416 (28.5)8714 (23.8)< 0.001
Patients filling 90-day prescriptions, n (%) 14,763 (40.4)13,310 (36.4)< 0.001

Patients in the diabetes cohort had 21.5% higher multivariate regression-adjusted odds of ≥ 80% adherence to OAB medications [odds ratio (OR) = 1.215; 95% confidence interval (CI), 1.169–1.263], during the 12-month evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2). Overall, the odds of adherence generally increased with age, and females had higher odds of adherence than men.

Figure 2.

 Adjusted odds and hazard ratios for study outcomes. p < 0.0001 difference between diabetes cohort and non-diabetes cohort for all outcomes

In the sensitivity analyses excluding patients with a 90-day OAB prescription, patients in the diabetes cohort had 16.4% higher odds of ≥ 80% adherence to OAB medications (OR = 1.164; 95% CI, 1.096–1.237) during the 12-month evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2).

Refill of a second overactive bladder medication prescription

A second OAB medication prescription was observed among approximately 69.7% of patients in the diabetes cohort compared with 64.5% of patients in the non-diabetes cohort (p < 0.001; Table 4). Patients in the diabetes cohort had 16.6% higher multivariate regression-adjusted odds of filling a second OAB medication prescription (OR = 1.166; 95% CI, 1.127–1.205) during the 12-month evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2).

In the sensitivity analyses excluding patients with a 90-day OAB prescription, patients in the diabetes cohort had 12.6% higher multivariate regression-adjusted odds of filling a second OAB medication prescription (OR = 1.126; 95% CI, 1.082–1.173) during the 12-month evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2).

Time to overactive bladder medication non-persistence

Discontinuation of OAB medications prior to the end of the 12-month evaluation period was observed among approximately 71.5% of patients in the diabetes cohort compared with 76.2% of patients in the non-diabetes cohort (p < 0.001). Mean time to discontinuation was 164.0 days (SD = 1.4 days) among patients in the diabetes cohort vs. 146.9 days (SD = 1.4 days) among patients in the non-diabetes cohort (p < 0.001; Table 3). Patients in the diabetes cohort had 10.3% lower multivariate regression adjusted hazard of non-persistence with OAB medications [hazard ratio (HR) = 0.897; 95% CI, 0.88–0.915] during the evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2).

In the sensitivity analyses excluding patients with a 90-day OAB prescription, patients in the diabetes cohort had 7.4% lower multivariate regression adjusted hazard of non-persistence with OAB medications (HR = 0.926; 95% CI, 0.907–0.946) during this evaluation period compared with those in the non-diabetes cohort (p < 0.0001; Figure 2).

Discussion

Many factors, including the impact of disease on the quality of life, disease severity, chronic nature of the disease and socioeconomic conditions, contribute to patients remaining adherent to their prescribed medications (11). Patients with several chronic diseases, such as diabetes, have well-established patterns of medication adherence (24,27), and may be more likely to adhere to therapy than patients with other chronic conditions, such as OAB (27). Based on these reports, we formulated our hypotheses that patients with OAB with a comorbid chronic disease such as diabetes mellitus would exhibit higher adherence and persistence rates compared with patients on treatment for OAB alone. Demonstration of increased adherence in such a chronic disease population could provide the basis for novel approaches to appropriate continuation of therapy in the overall OAB population.

A very large, geographically diverse sample of 36,560 patients with diabetes, who were initiating OAB medication, and a demographically matched comparison cohort of 36,560 patients without diabetes initiating OAB medication were evaluated for persistence and adherence to OAB medications during the 12-month period following initiation of these medications. Our analysis showed that female patients in the diabetes cohort had a lower mean number of visits to gynaecologists than those in the non-diabetes cohort (0.29 vs. 0.37; p < 0.001). A potential explanation for this difference is that patients in the diabetes cohort may have had their bladder symptoms managed by their primary-care physicians, endocrinologists or other specialists who they saw on a comparatively more frequent basis than the patients in the non-diabetes cohort. We also found that patients in the diabetes cohort had more cardiovascular comorbidities (Table 2), higher overall healthcare expenditure, hospitalisations, and office visits with healthcare providers (Table 3); therefore, it is logical to hypothesise that higher numbers of interactions with healthcare providers are essential to reinforce the importance of adherence to a medical regimen in this setting.

Compared with patients without diabetes, patients with diabetes had 21.5% higher odds of ≥ 80% adherence to OAB medications, 16.6% higher odds of filling a second OAB medication prescription and 10.3% lower hazard of non-persistence with OAB medications. A previously published study demonstrated that patients with diabetes are more likely to adhere to their medications vs. those with other chronic conditions (19). We speculate that, because patients with diabetes are accustomed to maintaining very specific medication regimens, have frequent contact with physicians, and are aware of the potential clinical impact of non-persistence and non-adherence to their medications, they may be more likely to adhere to their treatment regimen, regardless of whether the medications are for diabetes or OAB.

Adherence rates in this retrospective analysis were lower than several long-term OAB clinical trials (28–31); however, these studies do not measure drug adherence directly because they report on patients who continue medication throughout the full treatment period, where built-in incentives and follow-up may artificially increase adherence. Real-world adherence rates for OAB medication decrease to nearly half of those reported in clinical trials and, after 1 year, the adherence rate drops to nearly one-third (21,22,32). The odds of adherence generally increased with age. This finding was supported by results from another database analysis, which showed that adherence rates in patients with chronic conditions such as diabetes can improve with increasing age (33).

The significant short- and long-term consequences of poorly managed diabetes are well known to healthcare providers; therefore, they are likely to monitor these patients closely and frequently. Although morbidity associated with OAB can be significant (6), it may be perceived as less emergent and severe than a chronic disease such as diabetes. Hence, the plans for follow-up after diagnosis and implementing a treatment plan are apt to be more reactive and less frequent. Given that under-reporting of OAB symptoms by patients is well known (34), a ‘set it and forget it’ management plan is not likely to be specifically reinforced by regular interactions with the healthcare system. The results of this study suggest that such an approach could contribute to lower medication adherence. More frequent proactive follow-up could result in better medication adherence and presumably improve OAB outcomes, although this would require confirmation with a specifically designed prospective study.

Several limitations of this study should be considered. Claims data have inherent properties that limit the ascertainment of causal relationships. Except for the diagnosis of OAB, clinical information was primarily gathered from ICD-9-CM diagnosis codes, which are recorded by physicians to support claims for reimbursement and are subject to classification error. The MarketScan databases comprise a non-probability convenience sample of individuals with employer-sponsored health insurance and therefore the findings of this study may not be generalisable to other groups such as the uninsured or those covered by Medicaid or other payers. Because of the large number of patients (= 73,120) in the study, statistical tests for differences in the distributions of variables between the two cohorts often reached a high level of statistical significance when in fact the differences may have been considered clinically less meaningful. Measures of persistence and adherence were based on pharmacy claims, but as pharmacy claims indicate only prescriptions being filled, it was not possible to determine whether patients actually took the medications. Furthermore, although in our multivariate analyses of adherence we focused on the odds of ≥ 80% adherence to OAB medications, a choice informed by prior literature, it is possible that an alternative adherence threshold could have been more clinically meaningful or appropriate. Future studies to inform the most clinically meaningful adherence thresholds for OAB medications would advance this area of research. In addition, although we labelled patients as non-adherent if they did not regularly fill prescriptions for OAB medications, we cannot rule out the possibility that some patients may have been taking their drugs pro re nata. It is important to note, however, that pro re nata administration of this class of medications is off-label. Finally, the severity of neither diabetes nor OAB could be assessed in this analysis because this information is beyond the scope of MarketScan databases.

Despite these limitations, our methodology has many strengths unique to this approach. These observational results reflect the experience of patients outside the limits of an interventional environment, thereby presenting insight into everyday clinical practice and experience. Rather than focusing on one centre or a limited population, this study used a large database representing the experience of thousands of patients prescribed OAB medications across the USA. Finally, multiple sensitivity analyses demonstrated that the findings were very robust. Excluding patients with 90-day prescriptions slightly reduces the magnitude of the association between diabetes status and the outcomes; however, the association remains and we can therefore rule out the potential explanation that the higher proportion of patients with diabetes filling such prescriptions is the sole driver of the observed association between diabetes status and the outcomes.

Conclusions

Patients with diabetes mellitus can often suffer from bladder-related complications. Poor adherence and persistence to medication in such patients can result in symptom relapse and high retreatment rates. This study’s observational results demonstrate that after direct matching on index year, age, gender, and geographical region, and multivariate adjustment for other important demographics and clinical characteristics, patients in the diabetes cohort were more adherent and persistent to OAB medications and had higher odds of filling a second OAB medication prescription than patients in the non-diabetes cohort. Patients in the diabetes cohort also had more interactions with their healthcare providers. Although this may partially account for these trends, a causal relationship cannot be established with these data alone. These findings may have been affected by imbalances between cohorts with respect to factors such as the severity of diabetes or OAB symptoms. Furthermore, differences in patient response to OAB medication may have affected patient adherence; therefore, further research is needed to identify factors responsible for these findings, including disease severity, patient behaviour and response to treatment with OAB medications. This study has specific application among patients with comorbid diabetes and OAB. Physicians managing these patients should be aware of the importance of this association for the purposes of diagnosis and management, and should reinforce the need for adherence and persistence with prescribed medications for these conditions.

Acknowledgements

This study was undertaken with a research grant from Astellas Pharma US, Inc., and GlaxoSmithKline. Editorial support, including writing assistance, was provided by Radhika Bhatia, PhD, a medical writer at Envision Scientific Solutions, and was funded by Astellas Pharma Global Development, Inc., and GlaxoSmithKline.

Author contributions

All authors were involved in concept and design, analysis and interpretation of data, drafting of article, critical revision of the publication for intellectual content and approval of the article for publication. Stephen Johnston, David Smith and Kathleen Wilson were also involved in the acquisition of data.

Ancillary